[drupal-support] PHP compression and cacheing

[Julian Bond]

I've done some further testing on this.
Apache/2.0.52 (Win32) mod_perl/1.99_12 Perl/v5.8.2 PHP/4.3.4,
Turck's mmCache running on Win XP SP2

Start with
zlib.output_compression = off in php.ini
Drupal cache off
Logout
Firefox and IE6 work fine.

Now turn Drupal cache on and log out
All pages display in both firefox and IE6.
Pages can be seen to be cached in the dbms. Using the firefox live http 
headers extension you can see either the 304 unmodified or
Content-Encoding: gzip headers. What may be missing is the
Vary: Accept-Encoding header

Now switch zlib.output_compression = on in php.ini and restart apache
new pages display fine in firefox and IE6. Cached pages display as 
binary in firefox and throw an error in IE6 "unknown content type".
New pages are served with no gzip. Cached pages are either 304 or show 
Content-Encoding: gzip.

Now get back to a page that is fresh so you can login. login and turn 
Drupal cache off. All pages now display fine in both Firefox and IE6. 
Returned headers show
Content-Encoding: gzip
Vary: Accept-Encoding

As a final test I went into bootstrap.inc and in drupal_page_header() I 
commented out the line
header('Content-Encoding: gzip');
and changed the browser gzip check so it always inflated the cached data 
and everything worked fine with cache enabled and php set to compress.

So it looks pretty clear to me that Drupal should not be attempting to 
gzip encode the returned cache page if php or apache is also gzip 
encoding everything. I don't really see the benefit of storing the 
cached data gzip encoded as even on vast site, there's not going to be 
that much html.

The short term solution is just to always inflate the cached page before 
the print statement. The longer term solution is to provide an admin 
switch so that people without access to php.ini, .htaccess and so on, 
can choose.

And/or test for zlib_get_coding_type()=='gzip' to see whether it's 
turned on and take appropriate action. However this last will not find 
mod_gzip if that's installed in apache. Here's one way of doing it. 
Around line 262 of bootstrap.inc

       // Determine if the browser accepts gzipped data.
       if ( zlib_get_coding_type() == 'gzip' || 
(@strpos($_SERVER['HTTP_ACCEPT_ENCODING'], 'gzip') === false && 
function_exists('gzencode')) ) {


-- 
Julian Bond Email&MSM: julian.bond at voidstar.com
Webmaster:                 http://www.ecademy.com/
Personal WebLog:          http://www.voidstar.com/
M: +44 (0)77 5907 2173      T: +44 (0)192 0412 433
S: callto://julian.bond/